Various optoelectronic applications, such as optical telecommunications and intra-chip interconnects, involve transmitting optical signals and converting them to electrical signals at high data rates. Systems for performing such transmission and conversion usually require photodetector devices compatible with the speed and bandwidth of the optical signal. Typically, these devices are PIN detectors—i.e. semiconductor devices including an intrinsic region sandwiched between a p-type region and an n-type region—that have frequency responses in the GHz range and utilize optical waveguides as a conduit for directing light to the intrinsic region of the photodetector.
Generally, an optical waveguide is a planar, rectangular, or cylindrical structure having a high-index core surrounded by a low-index cladding. Light rays are predominantly confined in the core by internal reflection at the core/cladding interface, with a small portion of the light propagating in the cladding as an evanescent wave. In order to form a high-speed waveguide-based photodetector, a lightwave traveling in the optical waveguide is coupled to the intrinsic region of the photodetector, where the light is converted to photon-generated carriers. The carriers then diffuse out to the electrodes, e.g. the p- and n-type regions of the PIN detector, producing an electrical signal (e.g., a photocurrent) that corresponds to the detected light. To reduce scattering and improve detection efficiency, the intrinsic region of a PIN photodetector should be in direct contact with or sufficiently close to either a waveguide's butt end or to its surface. In the latter case, light can be coupled to the intrinsic region via the evanescent wave, a phenomenon referred to as “evanescent coupling.”
The speed of the detector generally depends on the time it takes for the photon-generated carriers to reach the electrodes. This time is referred to as the “transit time.” The narrower the intrinsic region, the shorter the transit time and the faster the detector. A fast photodetector allows for the detection and processing of high-speed optical signals.
As mentioned above, one potential application of such integrated photodetector devices relates to optical intra-chip interconnects. Recently, the increasing computing power of modern microelectronic devices has given rise to the need for smaller transistors and increased operating speeds that translate to higher density metallic interconnect lines carrying more current. The smaller cross-sectional dimensions of interconnects, however, generally lead to resistance-capacitance time delays and electromigration failure issues. Furthermore, the clock distribution typically consumes as much as 70% of the total power dissipated on a chip. Thus, conventional interconnects may impede further developments in microelectronic technology. In that regard, integrated photodetector devices present a promising alternative to the metallic interconnects, enabling novel microprocessor architectures by permitting significant increases in the intra- and inter-chip data transfer rates and reducing power consumption.
Implementation of these integrated photodetectors as optical interconnects for microelectronic devices, however, presents a number of challenges because different fabrication techniques are typically employed for microelectronic and optical components. Thus, it is desirable to incorporate the optical components onto a chip without jeopardizing CMOS processing compatibility or interfering with the operation of CMOS devices. In that regard, silicon-on-insulator (“SOI”) substrates, commonly used in CMOS fabrication, are particularly attractive for incorporating optical components with microelectronic devices. Specifically, using SOI as a starting substrate, low-loss waveguides can be defined in the top silicon layer. The high retractive index contrast between silicon and its oxide enables high light confinement in microscale strip waveguides. With respect to the photodetector, however, many applications having optical components that operate at wavelengths longer than silicon's absorption edge (1.3 μm to 1.55 μm) require an active material other than pure silicon to achieve acceptable absorption levels and generate photocurrent. Bulk germanium and/or silicon-germanium compounds have been previously proposed as suitable candidates for the photodetector material in silicon-based integrated waveguide photodetectors. See, for example, “Integration of Germanium Waveguide Photodetectors for Optical Intra-Chip Interconnects” by Rouvière et al. (Proceedings of SPIE, vol. 5453, 2004) and “Strain-Balanced Silicon-Germanium Materials for Near IR Photodetection in Silicon-Based Optical Interconnects” by Giovane (Ph.D. Thesis, MIT, 1998), both incorporated herein by reference.
Known approaches for integrating germanium-based photodetectors with silicon-based waveguides, however, suffer from several process- and performance-related drawbacks, including poor confinement and coupling efficiency, suboptimal integration density, and complexity of fabrication. For example, some techniques known in the art employ a vertical arrangement of the components, whereby a waveguide is disposed over the photodetector. Fabrication of these structures typically entails a surface-preparation step, e.g. chemical-mechanical polishing, prior to photodetector deposition.
Thus, there is a need in the art for versatile and cost-effective methods of fabricating integrated waveguide-based photodetector devices that are generally compatible with CMOS processing techniques. There is also a need in the art for integrated high-speed waveguide-based photodetector device with improved detection capabilities and integration density.